The authors describe a prototype intelligent information retrieval system based on using natural language understanding to efficiently locate captioned data. Features of this system are described in detail in the five sections of the paper. The problem of getting free-form information about images using suitable captions is covered in the first section. Several difficulties arising from the inherent order sensitivity of natural languages are pointed out, and other related work is presented here. Next, the authors’ approach in implementing the query-processing flow is described, and a series of developments in natural-language processing necessary for accurate information retrieval is presented. Conclusions about the corresponding retrieval effectiveness and search efficiency in terms of the evaluation criteria introduced by Salton and McGill are provided in the final sections of the paper.
The use of natural languages as information processing tools has recently become a topic of major interest and a very challenging research field. This work undoubtedly has merit in the area of retrieval of images. As is reported here, several difficulties are overcome by providing solutions to a series of subproblems, so further improvements and developments are expected in this area.